A force-based gradient descent method for $\text{ab initio}$ atomic structure relaxation

Abstract

Force-based algorithms for $\text{ab initio}$ atomic structure relaxation, such as conjugate gradient methods, usually get stuck in the line minimization processes along search directions, where expensive $\text{ab initio}$ calculations are triggered frequently to test trial positions before locating the next iterate. We present a force-based gradient descent method, WANBB, that circumvents the deficiency. At each iteration, WANBB enters the line minimization process with a trial stepsize capturing the local curvature of the energy surface. The exit is controlled by an unrestrictive criterion that tends to accept early trials. These two ingredients streamline the line minimization process in WANBB. The numerical simulations on nearly 80 systems with good universality demonstrate the considerable compression of WANBB on the cost for the unaccepted trials compared with conjugate gradient methods. We also observe across the board significant and universal speedups as well as the superior robustness of WANBB over several widely used methods. The latter point is theoretically established. The implementation of WANBB is pretty simple, in that no a priori physical knowledge is required and only two parameters are present without tuning.

arXiv comment8 pages, 9 figures

References in corpus (13)

• Crystal Structure Prediction via Particle Swarm Optimization
• Generation of Projector Augmented-Wave atomic data: a 71 elements validated table in the XML format
• Global minimum determination of the Born-Oppenheimer surface within density functional theory
• A universal preconditioner for simulating condensed phase materials
• Genetic Design of Enhanced Valley Splitting towards a Spin Qubit in Silicon
• Parallelizable Algorithms for Optimization Problems with Orthogonality Constraints
• Applicability of Kerker preconditioning scheme to the self-consistent density functional theory calculations of inhomogeneous systems
• On the existence of the optimal order for wavefunction extrapolation in Born-Oppenheimer molecular dynamics
• Model Hessian for accelerating first-principles structure optimizations
• Linear scaling relaxation of the atomic positions in nano structures
• SGO: An ultrafast engine for ab initio atomic structure global optimization by differential evolution
• Parallel 3-dim fast Fourier transforms with load balancing of the plane waves
• A curved line search algorithm for atomic structure relaxation

Cited by in corpus (1)

• Projected gradient descent algorithm for $\text{ab initio}$ crystal structure relaxation under a fixed unit cell volume